rising power of the network user - eucnc · carrier network architecture: metro-access segment...
TRANSCRIPT
Rising Power of the Network UserBiswanath (Bis) Mukherjee
Distinguished Professor of Computer Science University of California, Davis
Keynote Talk: Presented at: European Conference on Networking and
Communications (EuCNC 2018) Ljubljana, Slovenia
June 20, 2018
End-to-End-Service/Network Latency in 5G Applications
2
http://www.ieee802.org/3/ca/public/meeting_archive/2017/09/powell_3ca_1a_0917.pdf (T. Pfeiffer et al., Sep. 2017)
3
Carrier Network Architecture: Metro-Access Segment (ACK: Bob Doverspike, AT&T, Springer Handbook on Optical Nets, 2018)
4
5
Carrier Network Architecture: Metro + Backbone (ACK: Bob Doverspike, AT&T, Springer Handbook on Optical Nets, 2018)
Outline
6
1. 5G Briefing
2. The Network User and Its Growing Influence
3. (Network) User Experience: • Application performance in the cloud era…
Using advanced network analytics
5G: Wireless-Optical Convergence
7
Early Work (2007): • Suman Sarkar, Sudhir Dixit, and Biswanath Mukherjee,
“Hybrid Wireless-Optical Broadband-Access Network (WOBAN): A Review of Relevant Challenges,” IEEE/OSA Journal of Lightwave Technology, vol. 25, no. 11, Nov. 2007.
Led to: • Fixed-mobile convergence, Wireless-wireline convergence,
FiWi, etc. • Underpinnings of 5G Network — particularly Fiber
Fronthaul.
5G Requirements
8
1000X Mobile Data Volumes
10-100X Connected Devices
1000X End-user Data Rates
10X Lower Latency
50% Mobility Increase
10X Battery Life for Low-
Power Devices
1000X Lower Energy Consumption
“Big Three” 5G Technologies
9
Big Three
Ultra Densification
Millimeter Wave
Massive Multiple-Input Multiple-Output (MIMO)
Ultra-Densification
10
• More active nodes (for data transmission). • Smaller and more concentrated coverage of base stations • Candidate Architecture:
✓ Cloud Radio Access Networks (CRAN).
CRAN Architecture
11
▪ DU Cloud
• Computing resources
▪ Fronthaul • Bandwidth resources
▪ Radio Access
• Radio resources
Massive MIMO
12
Illustration of beamforming transmission in a massive MIMO system with 10 user terminals and an access point with 100 antenna elements. Source: LiU-ISY
mmWave
13
Recent Relevant Works (from my lab)
14
• Divya Chitimalla, Koteswararao Kondepu, Luca Valcarenghi, Massimo Tornatore, and Biswanath Mukherjee, “5G Fronthaul–Latency and Jitter Studies of CPRI Over Ethernet,” IEEE Journal of Optical Communications & Networking, vol. 9, no. 2, pp. 172-182, February 2017. (2018 Charles Kao Award for Best Paper in JOCN.)
• Yu Wu, Massimo Tornatore, Yongli Zhao, and Biswanath Mukherjee, “TDM EPON Fronthaul Upstream Capacity Improvement Via Traffic Classification and Sifting,” IEEE Globecom 2017. (TAOS Best Paper for IEEE Globecom 2017 ONS Symposium.)
Outline
15
1. 5G Briefing
2. The Network User and Its Growing Influence
3. (Network) User Experience: • Application performance in the cloud ere…
Using advanced network analytics
Networking R&D Trends
2000 201520102005
Backbone Networks(Carriers, network operators)
Cloud Networks (Service providers)
Application centric networking(Service providers)
User experience (Network users)
16
Backbone Networks
Consumers (You, I)
Enterprises (UCD, FDA)
Cloud service providers (Google, AWS, Microsoft)
Carriers (ATT, Verizon)
Traditional view Evolution towards cloud
17
Network Path Protection
Primary path
Backup path
Backup path
Protection against disaster by backup paths
Backbone networks must be resilient against any failures and disruptions!
18
Client
Protection by a backup path to DCProtection by a backup path to secondary DC
Protection by a backup path to disaster zone-disjoint secondary DC
Protection of both data (content) and connection against disasters
DC Z
DC X
DC Y
Cloud Network ProtectionContent/service protection in cloud networks ->
Content placement, routing, and protection of paths and content
Users are mainly interested in accessing data and services - irrespective of their physical locations
Content-centric/Application-centric networking!
19
20
CloudUser Equipment and Applications
Cloud DC
IT resources at the edge of networks
Transport Network
Access Network
Edge DCCoreNetworks
AccessNetworks
Access Networks
Cloud DC
User Equipment
Edge DCEdge DC
Edge-Cloud Side:Distributed DCs
User-Edge Side: Latency Sensitive
Cloud/Edge Computing
Network Users
21
!Users’ expectations!
Users are not patient anymore…
They require prompt services, or else they move to alternate
sites/apps
Era of “Data Deluge”
22
How to optimize each latency parts?
Self-driving Cars
Cloud
Access Networks
MEC Server
Video Streaming
Transport Networks
App latency
Access Net latency
Transport Net latency
App latency
Apps need to be more user-oriented should perform based on user’s device and network limitations
Application- and User-Centric Networking
How to optimize end-to-end latency?Improve and secure end-to-end services for the network users,
not only network infrastructures
Outline
23
1. 5G Briefing
2. The Network User and Its Growing Influence
3. (Network) User Experience:• Application performance in the cloud era…
Using advanced network analytics
24
Data Center
MainOffice
Branch Office
DNS LDAP
Data Center
SaaS
Challenge | Application-Centric Infrastructure
User Group 1
Application hosted in SDDC
Cloud-virtualized network services
3rd Party Applications
UserGroup 2
SD-WAN
UserGroup 3
Ennetix Confidential and Proprietary, 2018 25
Critical app not accessible, multiple days to root-cause
Impact: Very high triage timesMultiple teams involved, guesswork in localizing problem
1. Current solutions address either App Layer or Network Layer
2. Current network-layer solutions address one service or device (e.g., only DNS or Router 1)
Enterprise Challenges
26
Data Center
Data Center
Current Solution #1 | APM: No Network Visibility
Application hosted in SDDCAPM focus:
27
MainOffice
Branch Office
SD-WAN
NPM focus: Network devices in office locations
Current Solution #2 | NPM: No App Visibility
28
Data Center
MainOffice
Branch Office
DNS LDAP
Data Center
SaaS
User Group 1
Application hosted in SDDC
Cloud-virtualized network services
3rd Party Applications
UserGroup 2
SD-WAN
UserGroup 3
Solution | App-Specific Automated Analytics
29
Progression of Analytics
Technology Challenge #1
30
• Automated discovery of app service topology
Solution - AI-based causal dependency graph
Technology Challenge #2
31
• Automated path-performance mapping
Solution - ML-based packet-train technology
Technology Challenge #3
32
• App-specific user experience analytics Solution - Multi-layer Contextual Correlation
Technology Challenge #4
33
• App-specific “volumetric” analytics
Solution - Multi-Dimensional Macroflow Clustering
Technology Challenge #5
34
• Automated anomaly detection and pattern recognition
Solution - AI-based inferencing using parametric regression, model-based health scoring, and random decision forest
Ennetix Confidential and Proprietary, 2017 35
Summary | User Experience: Desirable Features
Capability How?
Analytics-driven visibility Network topology, App service topology, Traffic clusters
Automated anomaly detection Adaptive baselines, Heat map, Events
End-user SLA Application-delivery path performance, User transaction response time
Forensics Current as well as historical
Backup
37